1/*-
2 * Copyright (c) 1989, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 */
33
34#include <sys/cdefs.h>
35__FBSDID("$FreeBSD$");
36
37#include "opt_inet6.h"
38#include "opt_kgssapi.h"
39
40#include <fs/nfs/nfsport.h>
41
42#include <rpc/rpc.h>
43#include <rpc/rpcsec_gss.h>
44
45#include <nfs/nfs_fha.h>
46#include <fs/nfsserver/nfs_fha_new.h>
47
48#include <security/mac/mac_framework.h>
49
50NFSDLOCKMUTEX;
51NFSV4ROOTLOCKMUTEX;
52struct nfsv4lock nfsd_suspend_lock;
53
54/*
55 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
56 */
57int newnfs_nfsv3_procid[NFS_V3NPROCS] = {
58	NFSPROC_NULL,
59	NFSPROC_GETATTR,
60	NFSPROC_SETATTR,
61	NFSPROC_NOOP,
62	NFSPROC_LOOKUP,
63	NFSPROC_READLINK,
64	NFSPROC_READ,
65	NFSPROC_NOOP,
66	NFSPROC_WRITE,
67	NFSPROC_CREATE,
68	NFSPROC_REMOVE,
69	NFSPROC_RENAME,
70	NFSPROC_LINK,
71	NFSPROC_SYMLINK,
72	NFSPROC_MKDIR,
73	NFSPROC_RMDIR,
74	NFSPROC_READDIR,
75	NFSPROC_FSSTAT,
76	NFSPROC_NOOP,
77	NFSPROC_NOOP,
78	NFSPROC_NOOP,
79	NFSPROC_NOOP,
80};
81
82
83SYSCTL_DECL(_vfs_nfsd);
84
85SVCPOOL		*nfsrvd_pool;
86
87static int	nfs_privport = 0;
88SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_RW,
89    &nfs_privport, 0,
90    "Only allow clients using a privileged port for NFSv2 and 3");
91
92static int	nfs_minvers = NFS_VER2;
93SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers, CTLFLAG_RW,
94    &nfs_minvers, 0, "The lowest version of NFS handled by the server");
95
96static int	nfs_maxvers = NFS_VER4;
97SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers, CTLFLAG_RW,
98    &nfs_maxvers, 0, "The highest version of NFS handled by the server");
99
100static int nfs_proc(struct nfsrv_descript *, u_int32_t, struct socket *,
101    u_int64_t, struct nfsrvcache **);
102
103extern u_long sb_max_adj;
104extern int newnfs_numnfsd;
105extern struct proc *nfsd_master_proc;
106
107/*
108 * NFS server system calls
109 */
110
111static void
112nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
113{
114	struct nfsrv_descript nd;
115	struct nfsrvcache *rp = NULL;
116	int cacherep, credflavor;
117
118	memset(&nd, 0, sizeof(nd));
119	if (rqst->rq_vers == NFS_VER2) {
120		if (rqst->rq_proc > NFSV2PROC_STATFS) {
121			svcerr_noproc(rqst);
122			svc_freereq(rqst);
123			goto out;
124		}
125		nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
126		nd.nd_flag = ND_NFSV2;
127	} else if (rqst->rq_vers == NFS_VER3) {
128		if (rqst->rq_proc >= NFS_V3NPROCS) {
129			svcerr_noproc(rqst);
130			svc_freereq(rqst);
131			goto out;
132		}
133		nd.nd_procnum = rqst->rq_proc;
134		nd.nd_flag = ND_NFSV3;
135	} else {
136		if (rqst->rq_proc != NFSPROC_NULL &&
137		    rqst->rq_proc != NFSV4PROC_COMPOUND) {
138			svcerr_noproc(rqst);
139			svc_freereq(rqst);
140			goto out;
141		}
142		nd.nd_procnum = rqst->rq_proc;
143		nd.nd_flag = ND_NFSV4;
144	}
145
146	/*
147	 * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
148	 * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
149	 * mounts.
150	 */
151	nd.nd_mrep = rqst->rq_args;
152	rqst->rq_args = NULL;
153	newnfs_realign(&nd.nd_mrep, M_WAITOK);
154	nd.nd_md = nd.nd_mrep;
155	nd.nd_dpos = mtod(nd.nd_md, caddr_t);
156	nd.nd_nam = svc_getrpccaller(rqst);
157	nd.nd_nam2 = rqst->rq_addr;
158	nd.nd_mreq = NULL;
159	nd.nd_cred = NULL;
160
161	if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
162		/* Check if source port is privileged */
163		u_short port;
164		struct sockaddr *nam = nd.nd_nam;
165		struct sockaddr_in *sin;
166
167		sin = (struct sockaddr_in *)nam;
168		/*
169		 * INET/INET6 - same code:
170		 *    sin_port and sin6_port are at same offset
171		 */
172		port = ntohs(sin->sin_port);
173		if (port >= IPPORT_RESERVED &&
174		    nd.nd_procnum != NFSPROC_NULL) {
175#ifdef INET6
176			char b6[INET6_ADDRSTRLEN];
177#if defined(KLD_MODULE)
178			/* Do not use ip6_sprintf: the nfs module should work without INET6. */
179#define	ip6_sprintf(buf, a)						\
180			(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x",	\
181			    (a)->s6_addr16[0], (a)->s6_addr16[1],	\
182			    (a)->s6_addr16[2], (a)->s6_addr16[3],	\
183			    (a)->s6_addr16[4], (a)->s6_addr16[5],	\
184			    (a)->s6_addr16[6], (a)->s6_addr16[7]),	\
185			    (buf))
186#endif
187#endif
188			printf("NFS request from unprivileged port (%s:%d)\n",
189#ifdef INET6
190			    sin->sin_family == AF_INET6 ?
191			    ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
192#if defined(KLD_MODULE)
193#undef ip6_sprintf
194#endif
195#endif
196			    inet_ntoa(sin->sin_addr), port);
197			svcerr_weakauth(rqst);
198			svc_freereq(rqst);
199			m_freem(nd.nd_mrep);
200			goto out;
201		}
202	}
203
204	if (nd.nd_procnum != NFSPROC_NULL) {
205		if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
206			svcerr_weakauth(rqst);
207			svc_freereq(rqst);
208			m_freem(nd.nd_mrep);
209			goto out;
210		}
211
212		/* Set the flag based on credflavor */
213		if (credflavor == RPCSEC_GSS_KRB5) {
214			nd.nd_flag |= ND_GSS;
215		} else if (credflavor == RPCSEC_GSS_KRB5I) {
216			nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
217		} else if (credflavor == RPCSEC_GSS_KRB5P) {
218			nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
219		} else if (credflavor != AUTH_SYS) {
220			svcerr_weakauth(rqst);
221			svc_freereq(rqst);
222			m_freem(nd.nd_mrep);
223			goto out;
224		}
225
226#ifdef MAC
227		mac_cred_associate_nfsd(nd.nd_cred);
228#endif
229		/*
230		 * Get a refcnt (shared lock) on nfsd_suspend_lock.
231		 * NFSSVC_SUSPENDNFSD will take an exclusive lock on
232		 * nfsd_suspend_lock to suspend these threads.
233		 * This must be done here, before the check of
234		 * nfsv4root exports by nfsvno_v4rootexport().
235		 */
236		NFSLOCKV4ROOTMUTEX();
237		nfsv4_getref(&nfsd_suspend_lock, NULL, NFSV4ROOTLOCKMUTEXPTR,
238		    NULL);
239		NFSUNLOCKV4ROOTMUTEX();
240
241		if ((nd.nd_flag & ND_NFSV4) != 0) {
242			nd.nd_repstat = nfsvno_v4rootexport(&nd);
243			if (nd.nd_repstat != 0) {
244				NFSLOCKV4ROOTMUTEX();
245				nfsv4_relref(&nfsd_suspend_lock);
246				NFSUNLOCKV4ROOTMUTEX();
247				svcerr_weakauth(rqst);
248				svc_freereq(rqst);
249				m_freem(nd.nd_mrep);
250				goto out;
251			}
252		}
253
254		cacherep = nfs_proc(&nd, rqst->rq_xid, xprt->xp_socket,
255		    xprt->xp_sockref, &rp);
256		NFSLOCKV4ROOTMUTEX();
257		nfsv4_relref(&nfsd_suspend_lock);
258		NFSUNLOCKV4ROOTMUTEX();
259	} else {
260		NFSMGET(nd.nd_mreq);
261		nd.nd_mreq->m_len = 0;
262		cacherep = RC_REPLY;
263	}
264	if (nd.nd_mrep != NULL)
265		m_freem(nd.nd_mrep);
266
267	if (nd.nd_cred != NULL)
268		crfree(nd.nd_cred);
269
270	if (cacherep == RC_DROPIT) {
271		if (nd.nd_mreq != NULL)
272			m_freem(nd.nd_mreq);
273		svc_freereq(rqst);
274		goto out;
275	}
276
277	if (nd.nd_mreq == NULL) {
278		svcerr_decode(rqst);
279		svc_freereq(rqst);
280		goto out;
281	}
282
283	if (nd.nd_repstat & NFSERR_AUTHERR) {
284		svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
285		if (nd.nd_mreq != NULL)
286			m_freem(nd.nd_mreq);
287	} else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
288		svcerr_systemerr(rqst);
289	}
290	if (rp != NULL)
291		nfsrvd_sentcache(rp, xprt->xp_socket, 0);
292	svc_freereq(rqst);
293
294out:
295	NFSEXITCODE(0);
296}
297
298/*
299 * Check the cache and, optionally, do the RPC.
300 * Return the appropriate cache response.
301 */
302static int
303nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, struct socket *so,
304    u_int64_t sockref, struct nfsrvcache **rpp)
305{
306	struct thread *td = curthread;
307	int cacherep = RC_DOIT, isdgram;
308
309	*rpp = NULL;
310	if (nd->nd_nam2 == NULL) {
311		nd->nd_flag |= ND_STREAMSOCK;
312		isdgram = 0;
313	} else {
314		isdgram = 1;
315	}
316
317	/*
318	 * Two cases:
319	 * 1 - For NFSv2 over UDP, if we are near our malloc/mget
320	 *     limit, just drop the request. There is no
321	 *     NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the
322	 *     client will timeout/retry over UDP in a little while.
323	 * 2 - nd_repstat == 0 && nd_mreq == NULL, which
324	 *     means a normal nfs rpc, so check the cache
325	 */
326	if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL &&
327	    nfsrv_mallocmget_limit()) {
328		cacherep = RC_DROPIT;
329	} else {
330		/*
331		 * For NFSv3, play it safe and assume that the client is
332		 * doing retries on the same TCP connection.
333		 */
334		if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) ==
335		    ND_STREAMSOCK)
336			nd->nd_flag |= ND_SAMETCPCONN;
337		nd->nd_retxid = xid;
338		nd->nd_tcpconntime = NFSD_MONOSEC;
339		nd->nd_sockref = sockref;
340		cacherep = nfsrvd_getcache(nd, so);
341	}
342
343	/*
344	 * Handle the request. There are three cases.
345	 * RC_DOIT - do the RPC
346	 * RC_REPLY - return the reply already created
347	 * RC_DROPIT - just throw the request away
348	 */
349	if (cacherep == RC_DOIT) {
350		nfsrvd_dorpc(nd, isdgram, td);
351		if (nd->nd_repstat == NFSERR_DONTREPLY)
352			cacherep = RC_DROPIT;
353		else
354			cacherep = RC_REPLY;
355		*rpp = nfsrvd_updatecache(nd, so);
356	}
357
358	NFSEXITCODE2(0, nd);
359	return (cacherep);
360}
361
362/*
363 * Adds a socket to the list for servicing by nfsds.
364 */
365int
366nfsrvd_addsock(struct file *fp)
367{
368	int siz;
369	struct socket *so;
370	int error = 0;
371	SVCXPRT *xprt;
372	static u_int64_t sockref = 0;
373
374	so = fp->f_data;
375
376	siz = sb_max_adj;
377	error = soreserve(so, siz, siz);
378	if (error)
379		goto out;
380
381	/*
382	 * Steal the socket from userland so that it doesn't close
383	 * unexpectedly.
384	 */
385	if (so->so_type == SOCK_DGRAM)
386		xprt = svc_dg_create(nfsrvd_pool, so, 0, 0);
387	else
388		xprt = svc_vc_create(nfsrvd_pool, so, 0, 0);
389	if (xprt) {
390		fp->f_ops = &badfileops;
391		fp->f_data = NULL;
392		xprt->xp_sockref = ++sockref;
393		if (nfs_minvers == NFS_VER2)
394			svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program,
395			    NULL);
396		if (nfs_minvers <= NFS_VER3 && nfs_maxvers >= NFS_VER3)
397			svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program,
398			    NULL);
399		if (nfs_maxvers >= NFS_VER4)
400			svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program,
401			    NULL);
402		SVC_RELEASE(xprt);
403	}
404
405out:
406	NFSEXITCODE(error);
407	return (error);
408}
409
410/*
411 * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
412 * until it is killed by a signal.
413 */
414int
415nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args)
416{
417	char principal[MAXHOSTNAMELEN + 5];
418	int error = 0;
419	bool_t ret2, ret3, ret4;
420
421	error = copyinstr(args->principal, principal, sizeof (principal),
422	    NULL);
423	if (error)
424		goto out;
425
426	/*
427	 * Only the first nfsd actually does any work. The RPC code
428	 * adds threads to it as needed. Any extra processes offered
429	 * by nfsd just exit. If nfsd is new enough, it will call us
430	 * once with a structure that specifies how many threads to
431	 * use.
432	 */
433	NFSD_LOCK();
434	if (newnfs_numnfsd == 0) {
435		newnfs_numnfsd++;
436
437		NFSD_UNLOCK();
438
439		/* An empty string implies AUTH_SYS only. */
440		if (principal[0] != '\0') {
441			ret2 = rpc_gss_set_svc_name_call(principal,
442			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER2);
443			ret3 = rpc_gss_set_svc_name_call(principal,
444			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER3);
445			ret4 = rpc_gss_set_svc_name_call(principal,
446			    "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER4);
447
448			if (!ret2 || !ret3 || !ret4)
449				printf("nfsd: can't register svc name\n");
450		}
451
452		nfsrvd_pool->sp_minthreads = args->minthreads;
453		nfsrvd_pool->sp_maxthreads = args->maxthreads;
454
455		svc_run(nfsrvd_pool);
456
457		if (principal[0] != '\0') {
458			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2);
459			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3);
460			rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4);
461		}
462
463		NFSD_LOCK();
464		newnfs_numnfsd--;
465		nfsrvd_init(1);
466	}
467	NFSD_UNLOCK();
468
469out:
470	NFSEXITCODE(error);
471	return (error);
472}
473
474/*
475 * Initialize the data structures for the server.
476 * Handshake with any new nfsds starting up to avoid any chance of
477 * corruption.
478 */
479void
480nfsrvd_init(int terminating)
481{
482
483	NFSD_LOCK_ASSERT();
484
485	if (terminating) {
486		nfsd_master_proc = NULL;
487		NFSD_UNLOCK();
488		svcpool_destroy(nfsrvd_pool);
489		nfsrvd_pool = NULL;
490		NFSD_LOCK();
491	}
492
493	NFSD_UNLOCK();
494
495	nfsrvd_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_nfsd));
496	nfsrvd_pool->sp_rcache = NULL;
497	nfsrvd_pool->sp_assign = fhanew_assign;
498	nfsrvd_pool->sp_done = fha_nd_complete;
499
500	NFSD_LOCK();
501}
502
503